Public Report. November 2007

(1)

CEIOPS-DOC-19/07

C

E

I

O

P

S

’

R

e

p

o

r

t

o

n

i

t

s

t

h

i

r

d

Q

u

a

n

t

i

t

a

t

i

v

e

I

m

p

a

c

t

S

t

u

d

y

(

Q

I

S

3

3 )

)

f

o

r

S

o

l

v

e

n

c

y

I

Public Report

November 2007

(2)

1 Executive Summary

The European Commission (EC) requested the Committee of European Insurance and Occupational Pensions Supervisors (CEIOPS) to advice on the development of a new risk oriented solvency system (Solvency II) to be applied to European insurance and reinsurance undertakings, both at solo and group level.

As part of this project, a series of quantitative impact studies (QIS) have been scheduled, QIS3 being the third of such studies, to test the implications and impact of the different alternatives under scrutiny.

QIS3 is a test, and it has to be approached in that context, as it is not a final proposal for the Solvency II framework nor does it intend to be.

QIS3 is running well from an administrative point of view.The QIS3 process and results form the basis for preparing the QIS4. Regarding the organisation of QIS4, the exercise will be run by the European Commission.

The goals of QIS3 were fourfold:

- First, to obtain further information about the practicability and suitability of the calculations involved, and the alternatives tested.

- Secondly, CEIOPS was looking for quantitative information about the possible impact on the balance sheets, and the amount of capital that might be needed, if the approach and the calibration set out in the QIS3 specification were to be adopted as the Solvency II standard.

- Thirdly, information about the suitability of the suggested calibrations for the calculation of the solvency capital requirement (SCR) and minimum capital requirement (MCR) was collected.

- Fourthly, the effect of applying the QIS3 specification to insurance groups was tested for the first time.

The report is structured in a comprehensive way that dedicates different chapters to the different areas under scrutiny.

• Participation and adequacy of data:

Participants were allowed to take part on a best efforts and approximate basis. They could focus on material issues, in order to stimulate participation. As a result a substantial number of European undertakings participated in QIS3.

Both the number of insurers and the number of participating countries increased in comparison to the preceding QIS: In total, 28 out of 30 EEA member states took part in the study. The total number of solo company respondents was 1027, i.e. an increase of almost exactly 100% over QIS2, which had 514 respondents.

(13)

Of these 1027 undertakings, 330 are in the life sector and 511 in the non-life sector. Only 28 entities are classified as pure reinsurers. 158 are respondents that provide data for both life and non-life business (composites).

With 422 small and 418 medium undertakings participating, there have been almost as many small undertakings (<100 million € in premiums in non-life, <1000 million € in provisions in life) as medium undertakings that responded to QIS3. 187 large undertakings (>1,000 million € in premiums for non-life, and >10,000 million in provisions for life) submitted their data.

Participation, with respect to market share, was almost equal in all three sectors (life, non-life and composites), and for most countries it covered more than 60 percent. These numbers in most cases strongly increased since the last study, which reflects the particular interest of the industry in the quantitative impact studies and eventually Solvency II, as well as a recognition of the importance of such exercises.

The data provided by participants provide a broad basis for discussion. Some areas received a lower or more controversial feedback (e.g. the treatment of concentration/counterparty risk, or the approach for equity risk based on the duration of liabilities), underlining the fact that challenges lie ahead for CEIOPS in future exercises, starting with QIS4.

• Suitability, practicability and reliability:

The Technical Specifications set out for QIS3 were generally well received, although a few participants noted that guidance was insufficient.

In general, it took participants between one and three person months to complete QIS3. While many participants considered their data to be fairly accurate and reliable, this view was not fully shared by some supervisors.

At a qualitative level, CEIOPS requested participants to give their feedback on the suitability of the different aspects of solvency calculation as laid out in the technical specifications. The answers revealed as a general pattern that the calculation of the SCR is in general the item that raises the highest priority expectations, followed by the MCR, the assessment of eligible capital and the technical provisions. Guidance is expected more than prescriptive rules. Expectations for simplifications in the underlying methodologies generally lie in between.

• Architecture of the Solvency II system:

Solvency II follows a total balance sheet approach, as it considers both the asset and the liability side, both of them being evaluated following a market

(14)

consistency principle. An insurance company’s balance sheet can be presented in a stylised manner:

Table 1: Stylised balance sheet

Summary balance sheet Assets Liabilities

Reinsurance Own funds

Technical provisions (Risk margin element) Investments

Technical provisions (best estimate element)

Other assets Other liabilities

Total Total

Whilst in the current regime, the solvency assessment is based on accounting figures that are generally based on the national accounting standards, which vary widely (from market value to book value) between Member States, the Solvency II directive proposal introduces a common valuation principle based on a market consistent valuation of assets and liabilities.

The solvency assessment in this model relies on a few simple steps:

- Technical provisions (best estimate element) represent the best estimate of the future cash flows that will be paid or received until all of the insurance commitments are fulfilled, discounted using a risk free yield curve.

- Technical provisions (risk margin element): as capital will indeed be required until all insurance commitments are fulfilled, the cost of ensuring that the capital needed for subsequent years will be available is computed and booked on the liability side as the risk margin element of the technical provisions.

- Solvency capital requirement (SCR): The various risks that can have a material impact on the undertaking’s financial position are modelled and combined to calculate the required capital. Only those risks that have a probability of occurrence of more than 0.5% in the next 12 months are retained in this assessment. This gives the required capital for the coming year.

- If the total value of available assets is less than the sum of the technical provisions, the SCR required capital for the following year, the margin needed to ensure availability of capital in the subsequent years, and the value of the other liabilities, then the firm does not meet its solvency requirement. In the opposite situation, the firm is meeting is solvency

(15)

The Solvency II system is based on two levels capital requirement, representing two levels of intervention. A solvency capital requirement (SCR) sets the required level of capital for a licensed entity, calibrated to cover at least a one in 200 year event (99.5% Value at Risk). A lower minimum capital requirement (MCR) serves as the threshold for ultimate supervisory intervention, including winding-up, thus making the ease, robustness and reliability of calculation of the MCR important features.

• Financial impact:

The QIS3 report includes a chapter on the financial impact for participating firms of the methodology proposed, including a comparison with the current solvency regime (Solvency I). There are in fact differences in the way Member States have implemented the current EU solvency regime, and the existing national standards that build on this regime.

We could summarize the impact of the proposed approach as follows:

- There is no significant overall change in terms of neither composition nor size of the balance sheet when comparing Solvency I with Solvency II at an European level, however there may be national variations.

- Technical provisions – best estimate plus risk margin – tend to decrease vis-à-vis the current technical provisions because the implicit prudence that exists in the current regime is removed, thereby increasing the available capital. The average ratio of Solvency II provisions compared to Solvency I provisions varies more between countries in the non-life sector (70%-100%, with significant variations in the different lines of business) than in the life sector (90%-102%).

- As for the MCR, the vast majority of firms (98%) would not need to raise additional capital to meet it.

- The QIS3 SCR solvency ratio, i.e. the ratio of the available capital (own funds) to the SCR capital requirement, is lower for most participating undertakings than the current solvency ratio. In the non life sector, most undertakings show a decrease in their solvency ratios based on the QIS3 calculations; in the life sector, the results are more ambiguous, with an increase or decrease of the solvency ratio, depending on the Member States. This is consistent with the general philosophy of Solvency II, which takes risks into account more explicitly than the current framework.

- The regime does not require extra capital in the European insurance market as a whole. However, there will be a redistribution process as a consequence of introducing a risk oriented system where capital requirements will be in line with the risks assumed by the undertaking and

(16)

undertakings, the available surplus (i.e. the excess of available capital over the SCR) would increase by more than 50%, whereas in 34% of undertakings the available surplus would decrease by more than 50%. In addition, 16% of undertakings would have to raise capital to meet their SCR.

• Assessment of assets and liabilities:

The asset side valuation principle proved less demanding: investments were generally valued at market value where available, or in accordance with IFRS. A number of approaches for valuing illiquid assets and other non-insurance liabilities were observed.

• Assessment of technical provisions:

Technical provisions are the statutory insurance liabilities with which undertakings will cover expected losses arising from its portfolio.

Solvency II introduces a split in technical provisions between two main components, a best estimate and a risk margin, in line with the outcome of the previous QIS, i.e. QIS2.

1. Best estimate:

The approach followed in the majority of countries was very similar to the one in QIS2.

A difficulty commonly reported for the evaluation of provisions and the extent to which adverse events would change the payment of future benefits, arose when policyholders have a right to surrender their contracts at any time, or other behaviour dependent options. Expected future policyholder behaviour has to be modelled twice: once under future normal conditions to assess the best estimate of technical provisions, and once under future highly stressed conditions. These two calculations were needed to draw up the reference balance sheet, along with the required capital to cover the modelled adverse events. This proved to be one of the main technical challenges reported.

- Life business:

In most countries, the assessment of best estimate provisions for life business (other than for options and guarantees) was made on a deterministic approach basis. In some countries, a number of firms valued life policy options and guarantees directly through the use of a stochastic model, and some firms also took account in these models of links between economic variables, crediting rates/bonuses and lapses. Other countries did not explain how their life firms had valued options and guarantees on life policies. It is not clear what assumptions

(17)

were generally made by firms about the take-up of options by life insurance policyholders.

Firms were asked in QIS3 to include the value of all future bonuses for with-profit policies, that are legally or contractually required to be paid, or that might reasonably be expected to be paid, under current market conditions, within the calculation of their technical provisions. There was little specific information given by firms about how they assessed rates of future bonuses for this purpose. Some firms said they assumed a constant rate of bonuses based on current bonus levels, while others said that bonus rates had been included in their stochastic model1_{. In some countries, the amounts of any ‘surplus funds’ that have not yet}

been made available for distribution to with-profit policyholders, and could be utilised to cover any future losses arising, were deducted from the provisions and were shown as part of the ‘own funds’ on the balance sheet.

- Linked business:

For unit-linked business, most firms took the unit liability as the starting point for assessing the provisions. Most firms then added the present value of their best estimate of the non-unit cash flows, which might include the non-invested element of future premiums, as well as anticipated management fees; and where relevant they also valued any options or guarantees on these policies. The best estimate value of these non-unit cash flows was often negative, and for many linked policies, the provision held was less than the current surrender value. - Non-life business:

For non-life business, the assessment of claim provisions generally involved the application through expert judgement of some statistical or actuarial technique applied to either paid or incurred claims, and sometimes with adjustments for claims inflation.

Premium provisions were often calculated from the standard unearned premium reserves (UPR) calculation in the current balance sheet. Some participants were

not able to calculate best estimate premium provisions. Instead they used proxies based on the current accounting, which were provided by the national supervisor. This option was appreciated by the participants.

1_{When assessing the capital requirements, i.e. the SCR and MCR, firms were then} permitted to take account of the potential changes in the level of future bonuses that might be made following adverse future events, e.g. a change in market interest rates, a reduction in equity values or an increase in mortality rates. In the case of an adverse event, the bonuses (which are non guaranteed benefits) given to policyholders could decrease. As these future bonuses are included in the technical provisions for their full “unstressed” amount, this must be compensated for in the calculation of a lower capital requirement.

(18)

- Health business:

For special health (similar to life) business, only a minority of participants applied simulation techniques to produce the best estimate. For many companies, one of the major practical difficulties for the assessment of provisions was the quantity of required data, especially for non-life business, along with the need for some quite sophisticated models.

2. Risk Margin:

CEIOPS provided participants with a helper tab that was broadly used for the assessment of the risk margin in the provisions. It did not avoid that for a number of entities, mainly small and medium, the methodology was still complicated and data demanding.

• Assessment of the Minimum Capital Requirement (MCR):

As stated in the directive proposal, the MCR corresponds to an amount of eligible basic own funds below which policyholders and beneficiaries are exposed to an unacceptable level of risk, and its breach will trigger ultimate supervisory action (withdrawal of licence).

A modular MCR was tested in QIS3, with two alternatives regarding the market risk module: a simple factor-based approach based on asset-side volume measures and a more sophisticated factor-based approach, also taking into account liabilities and durations. As additional quantitative information, the CEA compact approach (MCR=33% SCR, either according to the standard formula or to the internal model) was calculated in the spreadsheet. According to the QIS3 results only 2-3% of undertakings would have to raise capital to meet their MCR. - Non-life MCR:

For non-life firms, the results for both MCR alternatives were just broadly consistent with the calibration target (80-90% Value at Risk over a one year time horizon).

- Life MCR:

For life and composite firms, the ratio of the MCR to the SCR shows a wide range of possible outcomes, including multiple instances of negative MCR/SCR ratios. The main driver for the problematic interaction with the SCR (and the negative ratios) for life and composite firms seems to be the methodology used to account for the loss absorbing capacity of future discretionary benefits.

• Assessment of the Solvency Capital Requirement (SCR):

The Solvency II directive proposal requires undertakings to hold eligible own funds to cover the SCR so that it covers unexpected losses derived from all

(19)

quantifiable risks that undertakings are exposed to, corresponding to the Value at Risk of the basic own funds of an undertaking, under a 99,5% confidence level over a one year period. The SCR will be calculated either by internal models or through a standard formula.

The QIS3 technical specifications laid out a modular approach for the SCR, combining the depicted risk types through correlation factors to a basic solvency capital requirement (BSCR).

The following risk modules (with submodules) were included in the SCR formula: 1. Market risk.

2. Life underwriting risk. 3. Non-life underwriting risk.

4. Health insurance underwriting risk. 5. Counterparty default risk.

6. Operational risk was taken into consideration at the top level.

Few comments were received on the overall modular approach for the SCR, which can be seen as an implicit approval of such an exercise where the participants mainly concentrate on the perceived flaws. Some countries even reported a general approval on the overall design.

In general, correlation coefficients as used in the SCR aggregation matrix were criticised only by a minority of participants, some indicating too prudent factors, others referring to the importance of tail correlations. The diversification benefits through correlation matrices were widely appreciated, however geographical diversification and the specific situation of niche operators were seen as areas for potential improvement.

Regarding the SCR composition for life firms, in most countries, market risk (before diversification) accounts for more than 70% of the Basic SCR (BSCR). Diversification effects of the overall aggregation of risk modules through the correlation matrix amounts to 20% on average.

For non-life firms, the respective underwriting risk composes the major part of BSCR in most countries, on average around 75%. Diversification effects are similar to those observed for life firms; however, variations in this figure are comparably smaller across countries.

For composite firms, diversification effects are largest, amounting to around 30%. In those undertakings, BSCR is mostly dominated by market risk.

Participants in several countries expressed their concerns with the methodology in QIS3 for the calculation of the adjustment for the risk mitigating effect (i.e. loss absorbency) of future profit sharing.

(20)

Liquidity risk in the insurance sector was seen as quite different from banking. A number of firms considered that principles related to liquidity risk should be a Pillar II issue only.

Many undertakings from several jurisdictions regretted the fact that for reasons of simplification, expected profit/loss in non-life business was no longer considered in the calculation as it was in QIS2, since this was considered to be an important contribution to the true economic valuation of non-life business. Inflation, liquidity, and credit risk for unearned commissions and other assets, were named as risks additionally to be taken account of in the SCR.

With respect to each risk module of the SCR standard formula:

1. Market risk:

The treatment of market risk was generally well received, and considered as a clear improvement over QIS2.

241 entities used the option to evaluate the effect of excluding free assets in the market risk module for their calculation of the SCR.

Concrete comments were made in relation to the different submodules that make up market risk.

¾ Currency risk submodule:

It was questioned whether a ‘one size fits all’ shock could be applied to currency risk, especially for currencies with fixed exchange rates. Nor was it deemed likely that all exchange rates move against the insurer with the same amount and in the same direction.

¾ Property risk submodule:

Main issues raised asked for an enhancement of granularity and some considered that the shock should differ in the different regions, while others considered this shock as being too high.

The alternative for property risk treatment based on the liability duration approach is controversial, with some countries strongly supporting it.

¾ Interest rate risk submodule:

This was seen by many participants as too simple for large undertakings and, at the same time, too complex for small ones.

¾ Equity risk submodule:

The equity risk submodule was altered considerably compared with QIS2: the equity shock was changed from a general 40% shock to a 32% shock for ‘global’

(21)

decreased from 0.75 to 0. This change in correlation yielded a reduction of market risk charge of on average 11% for life and 6% for non-life firms. Firms in some countries commented that they considered the revised factor and correlations too low in the context of observed market experience.

Some country reports noted that participants considered the equity risk module to be simplistic, for instance when compared with the interest rate risk module. It was suggested that the granularity for the equity risk module could be increased by increasing the number of indices, categorising them based on asset class, region and/or sector. Hedge funds were commented to be over penalised by being placed in the ‘other’ index, while the rationale to treat investments in participations similarly to other investments was questioned.

The alternative for equity risk treatment based on the liability duration approach is controversial, with some countries strongly supporting it and others not agreeing with the method.

¾ Spread risk submodule:

Participants from some countries requested that all credit risk (sub) modules be integrated into one module, as in QIS2. Also details of treatment for several asset classes were questioned. Treatment of government bonds, of unrated entities and considerations on the amount of the charge were raised by some participants.

¾ Concentration risk:

Some participants reflected on the idea of including geographical and sectoral concentrations within concentration risk. Simplification for the module was requested by other participants.

2. Life underwriting risk:

A scenario based approach was tested in the QIS3 for life underwriting risk, and subrisks were aggregated through a correlation matrix, in order to allow for the recognition of diversification effects.

Simplified approaches were also proposed for those entities not able to use the scenario based approach.

Catastrophe (CAT) risk in life was tested (risk of mass surrender of unit-linked contracts). The level of the factor used (75% of contract surrenders) was considered as too high by many participants.

3. Non-life underwriting risk:

Whilst for life a scenario based approach was followed, for non-life it was decided to follow factor based approaches.

(22)

CAT risk was included by aggregating scenario based CAT net costs, with some scenarios defined at a European level, and others left to local supervisors to define. Subjectivity of the selected scenarios under a 1 in 200 probability of occurrence (either by being more remote or more frequent) was raised by participants, who also reflected on potential overlaps or on the inappropriateness of the correlations used to combine CAT scenarios.

Removal of expected profit and loss (that were included in QIS2) was seen as a downside by many participants.

Non-life underwriting results were rated as excessive when compared to internal models results in a few countries, with some countries explaining it by a rise in the correlations that more than offset a decrease in assumed volatilities in the lines of business (LoB) since QIS2.

4. Health underwriting risk:

For health risk a separate module was tested that was applicable only to countries where the health system closely mimics the typical characteristics of life insurance. In addition, as compared to QIS2, the non-life line of business was split into three, in order to take into account e.g. of short term health insurance and workers compensation. The special health module appeared relevant for two countries: one country observed a general increase in the solvency ratio, another one reported that it either remains stable or increases.

Other countries classified the health business in the non-life underwriting risk module. For one country the non-life module did not adequately capture the risk mitigating effects of the national equalization system, thereby leading to overly severe solvency requirements.

5. Counterparty default risk (CDR):

In the counterparty default risk, the method of calculating the replacement cost was considered to be unclear. It was also questioned why the counterparty default risk module did not allow for recoveries after default. Participants of some countries requested a simplification in the CDR module.

6. Operational risk:

Operational risk in the QIS3 specifications was added to the BSCR as a separate module at the top level. The majority of undertakings seem to recognise operational risk as an area that requires special attention. However, many participants considered the operational risk module as tested under QIS3 as being too simplistic.

(23)

The comments by participants focussed on three points:

- First, participants opposed the 100% correlation between operational risk and other risk factors and demanded the recognition of diversification effects. This is a direct consequence of adding operational risk at the top level.

- Second, they criticised the module for not taking into account the quality of operational risk management within the insurance firm – in its current form, the formula would not give a sufficient incentive for the development of adequate risk management systems.

- As a third area of concern, participants mentioned the use of premiums and provisions instead of administrative costs – especially for unit-linked business the latter is seen as the more appropriate measure which would also be more in line with Basel II provisions.

CEIOPS also requested qualitative answers on the operational risk policy applied by participants. Risk management systems for operational risk differ significantly in their degree of sophistication. Large firms especially seem to have established strategies and procedures earlier than smaller firms.

• Own funds classification:

CEIOPS requested information on the type of capital (own funds) held by participants, classified following a three tier structure according to the fulfilment of a series of characteristics indicated in the Solvency II directive proposal (subordination, loss-absorbency, permanence, perpetuality, and absence of mandatory servicing costs). This approach is in line with the one used in banking, thus increasing cross-sectoral convergence.

Concerning the composition of the eligible capital elements covering the SCR, QIS3 took a principles-based approach, asking firms to classify their capital elements without providing concrete guidance. Consequently about 95 percent of capital was classified as Tier 1, including most subordinated debt instruments. For QIS4 purposes, this has been detected as an area that demands further concrete guidance to proceed.

In most countries, more than 50% of participating firms indicated having only Tier 1 capital, comprising primarily paid-up equity, retained earnings and valuation differences, and in some countries included the amounts of any ‘surplus funds’ that have not yet been made available for distribution to with-profit policyholders, and could be utilised to cover any future losses arising. The average proportion of Tier 1 capital across the industry was over 94% for both life and non-life firms in almost every country. Tier 2 and Tier 3 capital comprised mainly subordinated liabilities, members’ calls and unpaid share capital. For those firms with at least some Tier 2 capital, the average proportion

(24)

at least some Tier 3 capital, the average proportion of Tier 3 capital was less than 20% for life firms, and less than 33% for non-life firms in almost every country.

Many firms commented that the interpretation of the requirements for categorising elements of capital and particularly for calculating eligible elements was not sufficiently detailed.

Where ‘surplus funds’ were included, this was often a substantial proportion of overall own funds. Adopting this approach in conjunction with the reduction for profit sharing imbedded in the SCR would have a material impact on firms’ reported solvency ratios.

• Internal models:

As the directive allows for the calculation of the SCR through a standard formula or with either full or partial internal models, testing internal models implementation and use is of core importance for the design of the system.

For the first time in the QIS exercises, internal models results were requested in QIS3, yet 13% by number of the participants provided internal model results. This could be due to the lack of internal models, but also reluctance to share them, or being at an early implementation stage. This is work in progress and will demand additional testing in further QIS exercises.

Partial internal models are, in most of the cases, developed for the equity, interest rate and property risk sub modules. In some cases, also the spread and currency risk sub modules are considered. There is a wide dispersion in the reported ratios of the internal model calculations to the standard formula SCR. The internal models generally produce higher partial SCR for credit risk module than the standard formula. The picture is less clear in operational risk. Overall, the internal models in non-life insurance produce significantly lower total SCR than the standard formula. The average reduction in total SCR is about 25 percent. The reduction seems to be largely due to the non-life underwriting risk capital component.

• Insurance groups:

The chapter on groups sets out the QIS3 results. It is the first quantitative impact study that lays particular emphasis on group solvency.

The main aim of this exercise was to gather figures and information in order to refine further the approach for groups under the Solvency II regime. In particular, the study aimed at having a first view on the impact of diversification at a group level when implementing the standard approach as specified in this QIS, and at

(25)

gathering some information on internal models, in particular to compare them to the standard formula both on a qualitative and quantitative perspective. The study aims also at gathering information on how groups currently deal with group specific risks and particularly operational risk.

The fact that group aspects of QIS3 needed first to have implemented the ‘solo’ specifications, and the very tight time schedule that was allocated to the QIS3, certainly explain, at least partially, that groups were not able to provide CEIOPS with comprehensive figures. In particular, only few groups answered the QIS3 part related to internal models.

Nevertheless, several lessons can be drawn from this study. First, taking as a reference the standard approach as specified in QIS3, the study confirms that diversification benefits within a group can vary both from their sources and their amount. The levels of diversification that were reported by groups can then widely vary from one group to another.

Regarding the amounts of surplus capital in groups within the framework of QIS3, no general trend can be foreseen at this stage. This issue is all the more delicate because the group available capital largely depends on the valuation of liabilities that is not yet stabilised at ‘solo’ level.

Under these conditions, the main findings of QIS3 are more qualitative than quantitative. Even if there is some room for improvement in the standard approach proposed in the QIS3 regarding the calculation of the ‘diversified’ group SCR, some questions raised did not receive conclusive answers. This is notably the case for transferability issues and group specific risks. At this stage, even if the participants globally recognise the existence of these issues, there are no readily available solutions that can be adopted as they are. Further reflections are necessary in the next steps towards the implementation of Solvency II.

• Areas for further work: key lessons learned and challenges ahead QIS3 provided a wealth of information on a wide range of elements of the current Solvency II proposals. For some elements, multiple approaches were tested with the aim of being able to choose among the approaches after its completion. Below a number of areas are identified where political guidance would be helpful before embarking upon QIS4. These relate to the MCR, equity risk, groups and taxes.

1. MCR:

QIS3 tested multiple approaches for the MCR. The choice between these approaches has political as well as technical aspects. The Solvency II directive

(26)

proposal remains open on this issue. A choice needs to be made between the option of the MCR being a stand-alone capital requirement and the option of taking the MCR as a percentage of the SCR (the so-called Compact Approach). If the MCR retained is a stand-alone requirement, a choice has to be made as to its specific design that would satisfy the Directive Proposal’s criteria (80-90% VaR, simplicity and auditability, safety net, absolute floor). In this context, QIS3 tested the so-called Modular approach. From this analysis it can be concluded that for the non-life business the Modular approach displays an interaction with the SCR that is acceptable from the perspective of having a robust supervisory ladder of intervention. For life undertakings, on the contrary, the interaction between the Modular MCR and SCR was highly erratic and too volatile to be practical, due to the calculation of reduction for profit-sharing. For QIS4 purposes an improved stand-alone approach will therefore be designed and tested.

2. Equity risk:

QIS3 also tested two alternative approaches for equity risk in the SCR. In response to QIS2 feedback, the default charge on equity risk was reduced in three ways: the headline shock was reduced from 40 to 32 percent, the assumed correlation between equity and interest rate risk was reduced from 0.75 to 0, and participants were given the option to exclude equity holdings from their solvency calculation insofar as they consisted of free assets (i.e. assets not covering technical provisions nor the SCR). According to the qualitative returns, these modifications were well-received. In quantitative terms QIS3 demonstrated that for life undertakings the three modifications to the default standard formula tested in QIS3 would jointly lead to a reduction of roughly one third in the overall SCR compared to the QIS2 specifications. The hypothecation of assets, to be designated as ‘free assets’, and the interaction of this adjustment with the ‘surplus funds’ approach, would need further consideration and guidance. In addition to that, an alternative “duration-based” proposal was tested where equity holdings were tagged to the liability structure of the undertaking with declining risk weights. Incoming comments on this duration approach were rather mixed. The duration approach resulted in capital charges that varied from 50 to 100 percent of the default approach for equity risk (which for the most strongly affected life undertakings would take off another 15-20 percent of the total SCR). A choice should now be made, informed by political guidance, as to the approach to be tested under QIS4.

3. Groups:

QIS3 also tested for group issues such as the likely size of diversification benefits. Unfortunately, the partial character of the returns did not allow CEIOPS to make meaningful inferences about the size of any such effects. To this extent a more

(27)

targeted effort will have to be undertaken under QIS4, and the industry associations involved have already expressed their full commitment to this exercise. A related question here concerns third-country diversification benefits. While the QIS3 exercise was focussed at diversification and consequent solvency effects within the EEA, many globally active insurance groups noted that significant diversification effects also stem from third country undertakings established outside the EEA. Subject to this benefits being realizable, this aspect should be taken into account when drafting QIS4 specifications. This would of course require negotiations with third countries on the preferred allocation of diversification benefits within and outside the EEA.

4. Taxation:

During QIS3, some raised the question as to how to deal with taxes under Solvency II as in practice this may strongly influence the comparability of results. It was argued by some that deferred taxes should be counted either as reducing technical provisions or as part of available capital, since under stressed conditions insurance and reinsurance undertakings would not need to pay taxes. QIS3 was neutral and agnostic with regard to any accounting or tax issues, but as this is an issue that exceeds its scope, a decision may need to be taken before going into QIS4.

(28)

2 Introduction

2.1 Disclaimer

This report sets out the results from the third quantitative impact study (QIS3) conducted by CEIOPS to underpin the advice given to the European Commission to support the development of a sound insurance solvency regime. This impact study was mainly designed to test the calibration and the main structure for groups exposed in previously published CEIOPS answers to Calls for Advice from the European Commission. As such, QIS3 is a test and not a proposal for the final Solvency II framework.

Further, whenever in this report a reference is made to a statement from a clear minority of national supervisors (e.g. a reference to ‘one supervisor’), this is done because CEIOPS feels it is important to retain as much information from the individual country reports as possible. When for any issue only the view of a minority of supervisors is given, this means that the other supervisors did not give an explicit view on this issue.

CEIOPS took the experiences from QIS2 into account and endeavoured to keep the technical specifications as precise as possible and tried to minimise any misunderstandings by offering a Q&A process that allowed replies to QIS3 requests within the time frame of one week. Nevertheless, both undertakings and national supervisors may still have used different interpretations, eventually to the detriment of the comparability of the results. This may also explain some of the dispersion between country data, a phenomenon also found at country level between participants.

The report on groups sets out the results from the field study within QIS3. In the series of quantitative impact studies it is the first time that a particular emphasis is laid on group solvency. Comparisons with earlier studies are, thus, not possible.

In addition to reporting to national supervisors, the groups were asked to voluntarily report to a centralised database, which has been set up for extracting necessary quantitative and qualitative answers from group submissions, starting from the individual group level.

The quantitative analysis of the group results has been conducted at two different levels: (1) an analysis of insurance groups that directly submitted to the central database and (2) an analysis of the country reports as back up of the information retrieved from the database. The main difference between the levels

(29)

power due to different depth of information available to CEIOPS. The separation into two different assessments was necessary because, for confidentiality reasons, several QIS3 participants and supervisors were not able to supply their detailed data.

For these particular reasons CEIOPS points out that there might be some caveats to keep in mind when interpreting the results. First, the number of participants and their allocation across Europe plays a central role, for instance, to detect cross-country commonalities or divergence. On the other hand, standardised responses to the qualitative questionnaire make an adequate assessment of undertaking-specific difficulties with the QIS3 approach rather difficult. Under these circumstances it may become difficult to attribute certain challenges to a particular type or category of insurance group.

Furthermore, conclusions very much depend on the quality and clarity of submissions. Testing three different methods for determining a group capital requirement requires the use of comparable input parameters, in particular a comparable scope of consolidation, to allow for valid results. CEIOPS is aware that the group spreadsheet heavily depends on the availability of solo results of each subsidiary and hence cannot be completed in parallel.

In order to refine the analysis of the impact of the new solvency regime, a fourth quantitative impact study has been scheduled to take place in Spring 2008.

2.2 Structure of the report

The European Commission (EC) requested the Committee of European Insurance and Occupational Pension Supervisors (CEIOPS) to advise on the development of a new solvency system (Solvency II) to be applied to European insurance and reinsurance undertakings, including groups of such undertakings. For this purpose, CEIOPS has been requested by the EC to acquire insight into the possible quantitative impact of this new solvency regime through a series of quantitative impact studies (QIS). The results of QIS form a key input in the general impact assessment carried out by the EC.

CEIOPS launched a first QIS (QIS1) in Autumn 2005, the results of which were received in February 2006. The exercise focused on testing the level of prudence in technical provisions under several hypotheses. In the summer of 2006 CEIOPS conducted a more comprehensive second impact study (QIS2), which covered both technical provisions and the calculation of the solvency capital requirement (SCR) and minimum capital requirement (MCR). QIS2 focused on the

(30)

methodology of the solvency requirements; the calibration of the parameters was to be tested in a next study.

Taking into account the results of the previous QIS, CEIOPS has developed a new exercise (QIS3) that was launched in April 2007. The results of QIS3 are being reported in the document at hand.

The goals of QIS3 were fourfold. Firstly, QIS3 aimed at collecting further information about the practicability and suitability of the calculations involved. Secondly, QIS3 aimed at receiving quantitative information about the possible impact on the balance sheets, and the amount of capital that might be needed, if the approach and the calibration set out in the QIS3 specification were to be adopted as the Solvency II standard. Thirdly, QIS3 aimed at obtaining information about the suitability of the suggested calibrations for the calculation of the SCR and MCR. Fourthly, QIS3 studied the impact of these new proposals for insurance groups.

In addition to further improving the design and calibration of the standard formula, QIS3 included the assessment of the eligible elements of capital, based on the innovative proposal by the EC at the time of launching the exercise. Finally, since the publication in July this year of the Solvency II Proposal for a Directive2_{, the results of QIS3 will be of particular importance in the negotiation}

of the Framework Directive.

In principle, the structure of this report follows the structure of the country reports filled in by the national supervisors. Additionally, the report essentially consists of two parts, i.e. the report on solo entities and the report on insurance groups. The report is also characterised by the integration of data collected in the central database with the data received trough the national supervisors in their single group report. Chapter 3 studies the scope of the exercise by presenting information on the participating undertakings and the number of undertakings able to give quantitative input on the various calculations of the technical provisions and the solvency requirements. The subsequent chapter treats general comments on practicability and reliability. It covers both the necessary investments of undertakings participating in QIS3 and the reliability of data provided. Chapter 5 discusses the potential financial impact on each type of insurance undertaking; the following section depicts the assessment of technical provisions. Chapters 7 and 8 analyse the MCR and the SCR in detail, followed by a short section on own funds. Operational risk, which was, after QIS2, separated from the Basic SCR is then analysed in the following chapter 10. Chapter 11 summarises the results provided and the arguments raised on the internal model approach. The following section treats two special issues which are of particular

(31)

importance but cannot be treated solely in one of the former sections: First the suitability of QIS3 specifications for small insurance firms is evaluated; the second sub-section reports on the health insurance in form of life insurance. Chapter 13 extensively treats the results on insurance groups and their results. Finally, chapter 14 points out the areas of further work based on the preceding information.

2.3 Methodology

The Quantitative Impact Study essentially serves two purposes: It is intended to provide the best possible overview of European insurance undertakings and their risk exposure under the framework of the QIS3 Technical Specifications, while supervisors commit themselves to presenting a balanced view, including the necessary local information that needs to be taken account of for interpreting certain results while maintaining the highest level of confidentiality and professional secrecy, as stipulated in the legal background of their work, such that no participating entity needs to fear any disadvantages.

These to some extent opposing objectives are ascertained by a three step approach to the analysis:

1. Assessment of individual entity results by the national supervisor. The submissions are also checked for potential errors and misunderstandings before the procession to next step of analysis.

2. Building of ratios and basic statistics regarding the distribution of the sample (percentiles, weighted average, standard deviation and number of entities included).

3. Final assessment and aggregation into a European report by CEIOPS.

In a first step, supervisors analysed the participants’ QIS3 submissions and checked them for potential errors and misunderstandings.

Then, an IT tool extracted structured information from the national databases, containing all data from the individual spreadsheets. These databases served as the basis for the analytical tables which were generated for the country reports. Further, the databases are used to run additional analyses on the data obtained in QIS3: Complementing the IT tool, analytical macros were applied to the national datasets to be used for additional analysis. These macros do not reveal any confidential information but produce only the aggregated results needed. The national results, as provided by the respective supervisors, were finally compiled by CEIOPS and analysed for similarities, differences and potential anomalies. Together with the qualitative remarks by the participants the various

(32)

For the group report, the assessment was approached in a slightly different manner, but in principle using the same methodology: The national reports remained, however, the groups, on a voluntary basis, were also encouraged to submit to a central database because their business is in most cases not restricted to one single country. The assessment is essentially the same but the central database was assumed to bear several advantages over the national reports, inter alia: (1) more companies in the sample and therefore fewer confidentiality problems, (2) possibility to compare similar groups from different jurisdictions, (3) facilitated assessment of inter-group commonalities and divergences.

(33)

3 Participation and adequacy of data provided

(34)

3.1 QIS3 participation

A substantial number of European undertakings participated in the third quantitative impact study. Both the number of insurers and the number of participating countries increased in comparison to the preceding QIS3_{. These}

countries include Bulgaria, Cyprus, Latvia, Slovakia and Greece4_{. In total, 28 out}

of 30 EEA member countries took part in the study.

Table 2 below summarises the results and Table 3 shows the relative change in participants in comparison to QIS2 and with respect to size class.

Non-life insurers are classified according to the following table:

size class gross written premiums (million €)

large > 1 000

medium 100 – 1 000

small < 100

Life insurers are classified according to the following table:

size class gross technical provisions (million €)

large > 10 000

medium 1 000 – 10 000

small < 1 000

Apart from non-life insurers and life insurers for which the classification above can be applied directly, there are reinsurers and composite direct insurers which write both non-life business and life business. For those entities, the size class was assigned on a discretionary basis in line with the set classification of non-life insurers and life insurers described above. For instance,

− a composite insurer who conducts medium non-life business and small life business was classified at least medium;

− a composite insurer who conducts medium non-life business and medium life business was classified medium or large.

3_{Light blue countries on the map are those that already participated in QIS2. The new} participants are marked in dark blue.

(35)

Table 2: Number of respondents

Type of undertaking Small Medium Large Total

Life undertakings 116 135 79 330 Non-life undertakings 254 194 63 511 Pure reinsurers 12 10 6 28 Composites 40 79 39 158 All respondents 422 418 187 1027 Mutuals thereof 118 99 34 251

Health undertakings thereof 16 30 10 56

Table 3: Relative growth in participation

Type of undertaking Small Medium Large Total

Life undertakings 152% 61% 49% 80% Non-life undertakings 185% 92% 37% 117% Pure reinsurers 140% 400% 0% 115% Composites 167% 139% 44% 111% All respondents 172% 90% 42% 100% Mutuals thereof 203% 94% 113% 137%

Health undertakings thereof 100% 173% 233% 150%

The total number of solo company respondents is 1027, i.e. an increase of almost exactly 100% over QIS2, which had 514 respondents. All 24 countries which participated in QIS2 reported rising number of participants in QIS3. Of these 1027 undertakings 330 are in the life sector and 511 in the non-life segment. Only 28 entities are classified as pure reinsurers. 158 are respondents that provide data for both life and non-life business (composites). With 422 and 418 respectively there have been almost as many small undertakings as medium undertakings that responded to QIS3. There are 187 large undertakings that submitted their data. Among all respondents there have been 251 mutuals and 56 health undertakings, whereby it has to be mentioned that responses from undertakings performing health business according to ‘life’ principles were limited to five countries (AT, BG, DE, LU and NL).

In addition to the solo company submission, 51 groups submitted group calculations to national supervisors. For detailed information see chapter 13.1. 29 groups, which need not necessarily be included in the number of 51 stated before, submitted to the central database.

(36)

Figure 2: Growth in absolute numbers of respondents5 0 20 40 60 80 100 120 140 160 180 c h a n g e in n u m b e r o f p a rt ic ip a n ts Small 70 165 7 25 Medium 51 93 8 46 Large 26 17 0 12

Life undertakings Non-life

undertakings Pure reinsurers Composites

From QIS2 to QIS3 the number of small undertakings that took part in the study increased considerably. With a surge of 172 percent, the participation far more than doubled. In absolute numbers, this was a change of 267 respondents. An increase of 90 percent for medium size undertakings shows that there have been almost twice as many participants in this category. In comparison to the other size classes the increase in the participation of large undertakings is rather modest. Nevertheless, it has to be mentioned that in many EEA countries insurers of this size do not exist or already took part in QIS2. With respect to the sector (not taking account of health), the number of non-life undertakings increased the most. There are more than twice as many participants as in QIS2. An increase of 117 percent means an additional 275 respondents. Composites and pure reinsurers showed a similar relative surge, however from a much smaller basis. The number of life insurers increased by 80 percent or 146 enterprises.

5_{Mind that for QIS2 not all figures added up correctly because for some undertakings} the size classification was unknown. As a result the total of QIS2 to respondents plus the sum of all changes is slightly higher than 1027, i.e. the number of QIS3